+#include "libavcodec/dvdata.h"
+#include "libavutil/intreadwrite.h"
+#include "dv.h"
+
+struct DVDemuxContext {
+ const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
+ AVFormatContext* fctx;
+ AVStream* vst;
+ AVStream* ast[4];
+ AVPacket audio_pkt[4];
+ uint8_t audio_buf[4][8192];
+ int ach;
+ int frames;
+ uint64_t abytes;
+};
+
+static inline uint16_t dv_audio_12to16(uint16_t sample)
+{
+ uint16_t shift, result;
+
+ sample = (sample < 0x800) ? sample : sample | 0xf000;
+ shift = (sample & 0xf00) >> 8;
+
+ if (shift < 0x2 || shift > 0xd) {
+ result = sample;
+ } else if (shift < 0x8) {
+ shift--;
+ result = (sample - (256 * shift)) << shift;
+ } else {
+ shift = 0xe - shift;
+ result = ((sample + ((256 * shift) + 1)) << shift) - 1;
+ }
+
+ return result;
+}
+
+/*
+ * This is the dumbest implementation of all -- it simply looks at
+ * a fixed offset and if pack isn't there -- fails. We might want
+ * to have a fallback mechanism for complete search of missing packs.
+ */
+static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
+{
+ int offs;
+
+ switch (t) {
+ case dv_audio_source:
+ offs = (80*6 + 80*16*3 + 3);
+ break;
+ case dv_audio_control:
+ offs = (80*6 + 80*16*4 + 3);
+ break;
+ case dv_video_control:
+ offs = (80*5 + 48 + 5);
+ break;
+ default:
+ return NULL;
+ }
+
+ return frame[offs] == t ? &frame[offs] : NULL;
+}
+
+/*
+ * There's a couple of assumptions being made here:
+ * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
+ * We can pass them upwards when ffmpeg will be ready to deal with them.
+ * 2. We don't do software emphasis.
+ * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
+ * are converted into 16bit linear ones.
+ */
+static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
+ const DVprofile *sys)
+{
+ int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
+ uint16_t lc, rc;
+ const uint8_t* as_pack;
+ uint8_t *pcm, ipcm;
+
+ as_pack = dv_extract_pack(frame, dv_audio_source);
+ if (!as_pack) /* No audio ? */
+ return 0;
+
+ smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
+ freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
+ quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
+
+ if (quant > 1)
+ return -1; /* unsupported quantization */
+
+ size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
+ half_ch = sys->difseg_size / 2;
+
+ /* We work with 720p frames split in half, thus even frames have
+ * channels 0,1 and odd 2,3. */
+ ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
+ pcm = ppcm[ipcm++];
+
+ /* for each DIF channel */
+ for (chan = 0; chan < sys->n_difchan; chan++) {
+ /* for each DIF segment */
+ for (i = 0; i < sys->difseg_size; i++) {
+ frame += 6 * 80; /* skip DIF segment header */
+ if (quant == 1 && i == half_ch) {
+ /* next stereo channel (12bit mode only) */
+ pcm = ppcm[ipcm++];
+ if (!pcm)
+ break;
+ }
+
+ /* for each AV sequence */
+ for (j = 0; j < 9; j++) {
+ for (d = 8; d < 80; d += 2) {
+ if (quant == 0) { /* 16bit quantization */
+ of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
+ if (of*2 >= size)
+ continue;
+
+ pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
+ pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
+ if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
+ pcm[of*2+1] = 0;
+ } else { /* 12bit quantization */
+ lc = ((uint16_t)frame[d] << 4) |
+ ((uint16_t)frame[d+2] >> 4);
+ rc = ((uint16_t)frame[d+1] << 4) |
+ ((uint16_t)frame[d+2] & 0x0f);
+ lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
+ rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
+
+ of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
+ if (of*2 >= size)
+ continue;
+
+ pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
+ pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
+ of = sys->audio_shuffle[i%half_ch+half_ch][j] +
+ (d - 8) / 3 * sys->audio_stride;
+ pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
+ pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
+ ++d;
+ }
+ }
+
+ frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
+ }
+ }
+
+ /* next stereo channel (50Mbps and 100Mbps only) */
+ pcm = ppcm[ipcm++];
+ if (!pcm)
+ break;
+ }
+
+ return size;
+}
+
+static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
+{
+ const uint8_t* as_pack;
+ int freq, stype, smpls, quant, i, ach;
+
+ as_pack = dv_extract_pack(frame, dv_audio_source);
+ if (!as_pack || !c->sys) { /* No audio ? */
+ c->ach = 0;
+ return 0;
+ }
+
+ smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
+ freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
+ stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
+ quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
+
+ /* note: ach counts PAIRS of channels (i.e. stereo channels) */
+ ach = ((int[4]){ 1, 0, 2, 4})[stype];
+ if (ach == 1 && quant && freq == 2)
+ ach = 2;
+
+ /* Dynamic handling of the audio streams in DV */
+ for (i = 0; i < ach; i++) {
+ if (!c->ast[i]) {
+ c->ast[i] = av_new_stream(c->fctx, 0);
+ if (!c->ast[i])
+ break;
+ av_set_pts_info(c->ast[i], 64, 1, 30000);
+ c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
+ c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
+
+ av_init_packet(&c->audio_pkt[i]);
+ c->audio_pkt[i].size = 0;
+ c->audio_pkt[i].data = c->audio_buf[i];
+ c->audio_pkt[i].stream_index = c->ast[i]->index;
+ c->audio_pkt[i].flags |= PKT_FLAG_KEY;
+ }
+ c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
+ c->ast[i]->codec->channels = 2;
+ c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
+ c->ast[i]->start_time = 0;
+ }
+ c->ach = i;
+
+ return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
+}
+
+static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
+{
+ const uint8_t* vsc_pack;
+ AVCodecContext* avctx;
+ int apt, is16_9;
+ int size = 0;
+
+ if (c->sys) {
+ avctx = c->vst->codec;
+
+ av_set_pts_info(c->vst, 64, c->sys->time_base.num,
+ c->sys->time_base.den);
+ avctx->time_base= c->sys->time_base;
+ if (!avctx->width){
+ avctx->width = c->sys->width;
+ avctx->height = c->sys->height;
+ }
+ avctx->pix_fmt = c->sys->pix_fmt;
+
+ /* finding out SAR is a little bit messy */
+ vsc_pack = dv_extract_pack(frame, dv_video_control);
+ apt = frame[4] & 0x07;
+ is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
+ (!apt && (vsc_pack[2] & 0x07) == 0x07)));
+ c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
+ avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
+ c->sys->time_base);
+ size = c->sys->frame_size;
+ }
+ return size;
+}
+
+/*
+ * The following 3 functions constitute our interface to the world
+ */
+
+DVDemuxContext* dv_init_demux(AVFormatContext *s)
+{
+ DVDemuxContext *c;
+
+ c = av_mallocz(sizeof(DVDemuxContext));
+ if (!c)
+ return NULL;
+
+ c->vst = av_new_stream(s, 0);
+ if (!c->vst) {
+ av_free(c);
+ return NULL;
+ }